Minor Description
When creating a struct column in Dataframe, the code that ran without problems in version 3.3.1 does not work in version 3.4.0.
Example
val testDF = Seq("a=b,c=d,d=f").toDF.withColumn("key_value", split('value, ",")).withColumn("map_entry", transform(col("key_value"), x => struct(split(x, "=").getItem(0), split(x, "=").getItem(1) ) ))
In 3.3.1
testDF.show()
+-----------+---------------+--------------------+
| value| key_value| map_entry|
+-----------+---------------+--------------------+
|a=b,c=d,d=f|[a=b, c=d, d=f]|[{a, b}, {c, d}, ...|
+-----------+---------------+--------------------+
testDF.printSchema()
root
|-- value: string (nullable = true)
|-- key_value: array (nullable = true)
| |-- element: string (containsNull = false)
|-- map_entry: array (nullable = true)
| |-- element: struct (containsNull = false)
| | |-- col1: string (nullable = true)
| | |-- col2: string (nullable = true)
In 3.4.0
org.apache.spark.sql.AnalysisException: [DATATYPE_MISMATCH.CREATE_NAMED_STRUCT_WITHOUT_FOLDABLE_STRING] Cannot resolve "struct(split(namedlambdavariable(), =, -1)[0], split(namedlambdavariable(), =, -1)[1])" due to data type mismatch: Only foldable `STRING` expressions are allowed to appear at odd position, but they are ["0", "1"].; 'Project [value#41, key_value#45, transform(key_value#45, lambdafunction(struct(0, split(lambda x_3#49, =, -1)[0], 1, split(lambda x_3#49, =, -1)[1]), lambda x_3#49, false)) AS map_entry#48] +- Project [value#41, split(value#41, ,, -1) AS key_value#45] +- LocalRelation [value#41] at org.apache.spark.sql.catalyst.analysis.package$AnalysisErrorAt.dataTypeMismatch(package.scala:73) at org.apache.spark.sql.catalyst.analysis.CheckAnalysis.$anonfun$checkAnalysis0$5(CheckAnalysis.scala:269) at org.apache.spark.sql.catalyst.analysis.CheckAnalysis.$anonfun$checkAnalysis0$5$adapted(CheckAnalysis.scala:256) at org.apache.spark.sql.catalyst.trees.TreeNode.foreachUp(TreeNode.scala:295) at org.apache.spark.sql.catalyst.trees.TreeNode.$anonfun$foreachUp$1(TreeNode.scala:294) at org.apache.spark.sql.catalyst.trees.TreeNode.$anonfun$foreachUp$1$adapted(TreeNode.scala:294) at scala.collection.Iterator.foreach(Iterator.scala:943) at scala.collection.Iterator.foreach$(Iterator.scala:943) at scala.collection.AbstractIterator.foreach(Iterator.scala:1431) at scala.collection.IterableLike.foreach(IterableLike.scala:74) at scala.collection.IterableLike.foreach$(IterableLike.scala:73) at scala.collection.AbstractIterable.foreach(Iterable.scala:56) at org.apache.spark.sql.catalyst.trees.TreeNode.foreachUp(TreeNode.scala:294) at org.apache.spark.sql.catalyst.trees.TreeNode.$anonfun$foreachUp$1(TreeNode.scala:294) at org.apache.spark.sql.catalyst.trees.TreeNode.$anonfun$foreachUp$1$adapted(TreeNode.scala:294) at scala.collection.Iterator.foreach(Iterator.scala:943) at scala.collection.Iterator.foreach$(Iterator.scala:943) at scala.collection.AbstractIterator.foreach(Iterator.scala:1431) ....
However, if you do an alias to struct elements, you can get the same result as the previous version.
val testDF = Seq("a=b,c=d,d=f").toDF.withColumn("key_value", split('value, ",")).withColumn("map_entry", transform(col("key_value"), x => struct(split(x, "=").getItem(0).as("col1") , split(x, "=").getItem(1).as("col2") ) ))